Implementation of a Transformer Monitoring

Solution Per IEEE C57.91-1995 Using an

Automation Controller

Walter A. Castillo
CenterPoint Energy

Rekha T. Jagaduri and Prasanna K. Muralimanohar

Schweitzer Engineering Laboratories, Inc.

© 2012 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained
for all other uses, in any current or future media, including reprinting/republishing this material
for advertising or promotional purposes, creating new collective works, for resale or
redistribution to servers or lists, or reuse of any copyrighted component of this work in other
works.

This paper was presented at the 65th Annual Conference for Protective Relay
Engineers, College Station, Texas, April 2–5, 2012, and can be accessed at:
http://dx.doi.org/10.1109/CPRE.2012.6201252.
 1

Implementation of a Transformer Monitoring

Solution Per IEEE C57.91-1995 Using an
Automation Controller
Walter A. Castillo, CenterPoint Energy
Rekha T. Jagaduri and Prasanna K. Muralimanohar, Schweitzer Engineering Laboratories, Inc.

Abstract—Distribution network operation periodically exceeds damage to the transformer, thereby reducing its life. Power
the loading capabilities of a substation power transformer. This transformer failures can also lead to unplanned outages, which
can result in accelerated insulation aging and, in some cases, are not economical for the energy provider or the end user. A
failure due to complete deterioration of the insulation. The
solution described in this paper allows users to monitor critical
power failure causes multiple concerns for an electric utility,
substation transformer assets with a comprehensive transformer ranging from the replacement cost, the lead time to acquire a
thermal model per IEEE C57.91-1995. An automation controller new transformer, the cost to schedule an outage and, in some
calculates the transformer loading capability rating. The cases, the environmental cleaning cost associated with an oil
substation operator may initiate control action based on a spill. As a result, it is imperative to continuously monitor
warning or alarm that the transformer is overloaded and in these assets to ensure the effective and reliable operation of
danger of excessive insulation aging or loss-of-life. The hottest-
spot temperature of the transformer core is used as a basis for
the electric power system.
calculating the insulation aging acceleration factor and loss-of-
life quantities. CenterPoint Energy intends to use this III. TRANSFORMER LOADING CAPABILITY
information to better manage their transformer assets and MONITORING METHODS
further utilize the power transformer near its actual thermal
Protective relays are vital components required for fault
limits.
detection and isolation of a transformer from a fault. In
I. INTRODUCTION addition to these relays, other monitoring techniques are
employed to adequately protect and monitor the transformer.
CenterPoint Energy in Texas has distribution substations Several transformer monitoring techniques have been
that deliver power to about 5 million customers. CenterPoint developed for power transformer diagnosis [2]. The first
Energy is experiencing steady load growth, similar to other technique is a time-based monitoring solution that involves
electric utilities across the United States. As the demand for performing various periodic offline tests to detect incipient
electric power increases, the load on each distribution problems. These offline tests can only be employed after a
transformer also increases. Increased loads on distribution transformer outage or during scheduled maintenance. This
transformers can lead to additional mechanical wear and method is expensive and labor intensive.
insulation deterioration of the transformer. Prolonged The present trend in the power industry is to move from
exposure to abnormal operating scenarios can also lead to time-based monitoring to a condition-based monitoring
transformer failures and distribution outages. Such failures system. The condition-based technique can supply information
may be prevented by using an effective real-time transformer about the transformer in real time and process the information
monitoring solution for all distribution power transformers. to determine any corrective actions that may be needed to
This paper outlines the need to monitor power transformers protect the transformer from overload. Condition-based
and discusses the methods used to monitor them. The paper monitoring contains a wide range of methods [3] [4] [5],
presents the design and algorithm involved in implementing a including detecting partial discharges and insulation
transformer monitoring system using an automation controller. degradation, diagnosing winding deformations, monitoring
Laboratory tests were conducted for multiple operating dissolved gas evolution, and assessing the thermal condition
scenarios, and the test results obtained are also presented. of the transformer. The solution specified in this paper entails
the use of an online monitoring solution based on assessing
II. NEED FOR POWER TRANSFORMER MONITORING the thermal condition of the transformer.
Large oil-immersed power transformers are the most The development of modern microprocessor-based relays
expensive components of a transmission and distribution with enhanced processing capacity enables relays to perform
system and are considered the most valuable assets of an additional monitoring tasks while providing fault protection.
electric power utility system. System abnormalities, excessive With this enhancement, protective relays now can perform
loading, switching, and ambient conditions contribute to both protection and monitoring functionalities. The thermal
accelerated aging and can lead to the sudden failure of model-based monitoring functions in protective relays are
transformers [1]. Some of these failures can cause irreversible based on comparing the obtained transformer internal
 2

temperatures and loss-of-life values with predefined limits. IV. SYSTEM OVERVIEW
The relay is usually programmed to issue a warning if these Fig. 1 illustrates the transformer monitoring solution for a
limits are exceeded. typical medium-voltage substation with two distribution
However, CenterPoint Energy needed a solution to provide transformers. The system design involves the use of data
a predictive loading assessment based on present loading acquisition (DAQ) modules for each distribution transformer.
conditions and ambient temperatures. In other words, it was The DAQ modules are configured to obtain the necessary data
essential to determine the loading capability of a transformer for monitoring their individual distribution transformers. An
for the following two hours, providing the operator sufficient automation controller is used to perform the calculations
time to roll over loads and dispatch field personnel in case of necessary for estimating the loading capability of the
an emergency. An automation controller was selected to transformer for the next two hours without exceeding the
perform these additional calculations. aging and thermal limits.
The internal temperatures and aging of a transformer are The DAQ modules are placed in standard National
calculated based on the equations given in IEEE C57.91-1995 Electrical Manufacturers Association (NEMA) 4X outdoor
[6]. The convergence process provided in IEEE C57.91-1995 enclosures in the substation yard next to the distribution
is modified to determine the loading capability for the next transformers. Resistance temperature detectors (RTDs) from
two hours.

Control Center Control

Building

Telco Digital Data

System Circuit IRIG-B Connection
Serial DNP3 Outstation
Data Service
DNP3 Serial Over Fiber
Unit Modem
Hard-Wired Connection
Transducer/Thermocouple
Ethernet Communication
Substation
Remote M Fan Motor
Terminal
Unit (RTU)
NEMA Enclosure NEMA Enclosure
in Yard Substation in Yard
Status Computer Status
Control Control

DAQ Automation DAQ

Module Controller Module

Ambient Top-Oil Top-Oil Ambient

Temperature Temperature Temperature Temperature
Global
Positioning
System
(GPS) Clock

Substation
12 kV Bus

M M M M M M M M M M M M

Cooling Bank Cooling Bank

Distribution Transformer 1 Distribution Transformer 2

Fig. 1. Overall Transformer Monitoring System Architecture

 3

the transformer and ambient temperature sensors are allocate and dispatch field personnel in the event that any of
connected to the DAQ module analog input cards. These the substation transformers trip or fail. In addition to the two-
RTDs are used to measure the top-oil and ambient hour rating, the algorithm also computes the temperature and
temperatures. In addition to these temperature analogs, the aging limits of the transformer reached at the end of the two-
DAQ module receives the distribution transformer load hour interval.
current and voltage from the existing instrument transformers The highest temperature point (hottest spot) in the
used in protection and metering circuits. transformer is a major factor that influences the aging of a
The DAQ modules are connected via a serial fiber-optic transformer. In addition, the insulation aging also depends on
communications channel to the automation controller, which the amount of moisture and oxygen in the transformer oil.
is located in the control building. The automation controller Modern oil preservation systems minimize the effect of
calculates the real-time transformer loading capability, moisture and oxygen present in the oil. They isolate the
insulation aging factor, and the hourly and daily loss-of-life hottest-spot temperature as the only major time-based factor in
values of the transformer. The substation computer is determining the insulation aging.
connected via an Ethernet local-area network (LAN) to the IEEE C57.91-1995 offers two methods of calculating the
automation controller. The substation computer runs a custom hottest-spot temperature. One method uses the transformer
human-machine interface (HMI) application that displays the bottom fluid temperature, which requires the heat-run test data
real-time status of all the transformers in the substation (see to have parameters such as the bottom fluid temperature at
Fig. 2). These data are transmitted to the utility control center rated load and the bottom fluid rise over ambient temperature
via the modem connected to the substation RTU to provide at rated load. These additional parameters may not be
remote monitoring capability for the power transformers at the available for most transformers that have been in service for
substations. The automation controller is, furthermore, 30 to 40 years. The second method uses the winding hottest-
connected to a GPS-synchronized clock to provide time- spot rise over ambient temperature and assumes that the oil
synchronized measurements to the control center. temperature in the cooling ducts is the same as the oil
temperature in the tank top oil for all operating conditions.
V. DETERMINATION OF LOADING CAPABILITY
A. Thermal Model
IEEE C57.91-1995 is used to determine the loading
IEEE C57.91-1995 assumes the hottest-spot temperature
capability of mineral-oil-immersed distribution and power
consists of three components: ambient temperature, top-oil
transformers [6]. This standard provides the technique to
rise over ambient temperature, and winding hottest-spot rise
calculate and determine the temperatures and aging factors of
over top-oil temperature. The hottest-spot temperature is
a transformer. It also presents a convergence algorithm that is
calculated as:
used to determine transformer loading capability at a
particular instant. θH = θA + ΔθTO + ΔθH (1)
For the CenterPoint Energy transformer monitoring where:
solution, it was vital to determine the permissible additional θH is the hottest-spot temperature.
load that could be applied to the transformer at the present
θA is the ambient temperature.
loading condition for a period of two hours without exceeding
the thermal limits (two-hour rating). In essence, this ΔθH is the hottest-spot rise over top-oil temperature.
calculation gives the operator time to roll over loads and ΔθTO is the top-oil rise over ambient temperature.

Fig. 2. Transformer Monitor Local HMI Screen

 4

The top-oil and ambient temperatures are obtained as controller. The flow chart depicting the convergence algorithm
measured values from the DAQ module. IEEE C57.91-1995 is shown in Fig. 3.
provides equations to determine the change in the ΔθH and
ΔθTO temperatures. The hottest-spot temperature determines
the transformer accelerated aging factor, FAA. FAA, in turn,
adds up over a period of time to the equivalent aging factor,
FEQA, and the daily and total loss-of-life values. The equations
for obtaining the change in temperatures follow.
The top-oil rise over ambient temperature is given by the
following two equations [6]:
−Δt
τTO
ΔθTO = (ΔθTO,U − ΔθTO,i ) • (1 − e ) + ΔθTO,i (2)
EXPn
⎛ K 2 • RATL + 1 ⎞
ΔθTO,U = ⎜ ⎟ • ΔθTO,R (3)
⎝ RATL + 1 ⎠
where:
ΔθTO,U is the ultimate top-oil rise over ambient
temperature in degrees Celsius.
ΔθTO,R is the top-oil rise over ambient temperature at rated
load in degrees Celsius.
ΔθTO,i is the initial top-oil rise over ambient temperature
in degrees Celsius.
RATL is the ratio of load loss at rated load to no-load
loss.
K is the ratio of present load to the rated load.
EXPn is an empirically derived exponent used to
calculate the variation of ΔθTO,U with changes in load.
τTO is the oil time constant of the transformer for any load
and for any specific temperature differential between the
ultimate top-oil rise and the initial top-oil rise temperature
in hours.
Δt is the duration of load in hours.
Equation (2) is the solution for the first order differential
equation shown by (4).
dΔθTO
τTO • = −ΔθTO + ΔθTO,U , ΔθTO (0) = ΔθTO,i (4)
dt
Using a forward Euler’s approximation of (4), the equation
to calculate the change in top-oil temperature from a previous
time interval is determined [7]. The time interval is chosen to
be two hours. Similar approximations are applied in
determining the hottest-spot temperature. With the calculated
hottest-spot temperature, the FAA value for transformers with
55°C average winding rise is determined using (5).
⎡15000 15000 ⎤
⎢ − ⎥
⎣ 368 θH + 273 ⎦
FAA = e (5)
With the temperature and aging factors calculated, the
automation controller can determine the loading capability of
the transformer.
B. Convergence Algorithm
A convergence algorithm to determine the precise loading Fig. 3. Convergence Algorithm
capability of the transformer without exceeding the
temperature and aging limits is programmed in the automation
 5

For every computational cycle, the automation controller substation transformer failure (Case Study 2). The transformer
gathers analog data from the DAQ modules. The limits for data and the heat-run data of the transformer under test are
temperatures and aging factors are set at the start of the cycle. shown in Table II and Table III, respectively.
A load multiplier is used to vary the value of K to determine TABLE II
the new temperatures and aging values as follows: TRANSFORMER DATA
K i = Load in per unit • LOAD MULTIPLIER i (6) Parameter Value
where: Capacity 15/20/25 MVA at 55°C
i is the iterative variable. High-Voltage Winding 67 kV
The algorithm starts the convergence process by setting the Low-Voltage Winding 12.47 kV
load multiplier to 1 and determining the present MVA on the
transformer. The base MVA is set based on the active Impedance 7.79%
transformer cooling stage. The active cooling stage is TABLE III
determined by predefined activation rules, such as those HEAT-RUN DATA
shown in Table I. The cooling stage selection is validated by Parameter Value
current detection of cooling bank loads by the DAQ modules.
Transformer Life 180,000 hours
TABLE I
(1)
COOLING STAGE ACTIVATION RULES Cooling Stage OA FA(2) FOA(3)

Criteria Threshold Cooling Top-Oil Rise Over Ambient Temperature

47.6 39.3 39.6
at Rated Load, °C
75°C Cooling Stage 1 Activation
Top-Oil Temperature Hottest-Spot Rise Over Top-Oil
17.4 25.7 25.4
85°C Cooling Stage 2 Activation Temperature at Rated Load, °C

Hottest-Spot 100°C Cooling Stage 1 Activation Winding Exponent, EXPn 0.8 0.8 0.9
Temperature 120°C Cooling Stage 2 Activation Oil Exponent, EXPm 0.8 0.8 1.0
1.00 pu Cooling Stage 1 Activation Top-Oil Time Constant at Rated Load
Load Current With Initial Top-Oil Rise of 0°C, τTO,R, 9.322 8.133 6.715
1.20 pu Cooling Stage 2 Activation Hours

With all the set points and parameters determined, the Winding Time Constant, Hours 0.1
algorithm begins the convergence loop. In every loop, the load (1)
Oil Air (OA) Stage 1 Cooling
multiplier is changed to vary K. For every new load multiplier, (2)
Forced Air (FA) Stage 2 Cooling
the algorithm determines the temperatures and aging factors (3)
Forced Oil-Air (FOA) Stage 3 Cooling
reached at the end of two hours. At the end of the convergence
The test setup consisted of a secondary injection of three-
loop, the algorithm fine-tunes the load multiplier value to
phase currents and voltages with a test set. This test set output
reflect the actual load that can be impressed on the transformer
was varied to simulate actual distribution transformer loading
without exceeding the temperature and aging limits.
patterns. The inputs from an RTD simulator were used to
represent the inputs obtained from an ambient temperature
VI. TEST RESULTS
probe. These laboratory tests used the calculated top-oil
The CenterPoint Energy transformer monitoring solution temperature to determine the hottest-spot temperature. In field
was tested in a laboratory using multiple operating scenarios. implementation, the top-oil temperature is a measured value
Some of these scenarios included a single transformer four- obtained from an RTD probe placed in a thermal well inside
day test run (Case Study 1) with an example distribution the transformer core. In the laboratory, the analog inputs from
transformer load profile from IEEE C57.91-1995 and a two- the test set and the RTD simulator were connected to the DAQ
transformer substation contingency test involving a single modules.
 6

A. Case Study 1: Four-Day Load Profile B. Case Study 2: Contingency Run

The single transformer four-day test run involved A contingency situation for a two-transformer substation
determining the response of the transformer monitoring involving the failure of one of the distribution transformers
solution for an example load profile from IEEE C57.91-1995. was simulated, and the response of the convergence algorithm
A constant ambient temperature of 34.5°C was maintained is shown in Fig. 5. A typical load profile from a CenterPoint
during the test, and the base MVA of the unit was set to Energy substation with two distribution transformers was
15 MVA (i.e., the transformer was run without additional selected. The time duration of the test was chosen to include
cooling). Fig. 4 shows the results obtained at the end of the the peak load and the highest ambient temperature. During the
test period. From these results, it is evident that the predicted time of the test, the ambient temperature was varied to
two-hour rating is dependent on the transformer load and the illustrate its impact on the two-hour rating.
top-oil and hottest-spot temperatures. In other words, as the When specific set points for the temperatures or the load
temperatures and present load on the transformer approach the current were reached, the algorithm automatically changed the
limit, the algorithm determines if load reduction is required base MVA based on the cooling stage selection rules shown in
depending on the cooling stages. In this test simulation, the Table I. The test started at 09:35, with both the substation
transformer was run without any additional cooling; hence the transformers being healthy. During this period, the
algorithm predicts the amount of load that should be removed temperatures and load were within the limits specified for
to prevent the transformer from aging more rapidly. Cooling Stage 1 (OA). At about 10:35, a substation event

Fig. 4. Results of Four-Day Test Run

Transformer Load (MVA)

Temperature (°C)

Fig. 5. Contingency Run

 7

caused one of the transformers to fail. At that time, the entire [6] IEEE Standard C57.91-1995, IEEE Guide for Loading Mineral-Oil-
Immersed Transformers, 1996.
load for that substation was transferred to the transformer
[7] B. C. Lesieutre, W. H. Hagman, and J. L. Kirtley Jr., “An Improved
under test. For the next five hours, the additional load
Transformer Top Oil Temperature Model for Use in an On-Line
impressed on the in-service transformer caused the Monitoring and Diagnostic System,” IEEE Transactions on Power
transformer to activate Cooling Stage 2 (FA). Fig. 5 shows the Delivery, Vol. 12, No. 1, January 1997, pp. 249–256.
predicted two-hour rating and the top-oil and hottest-spot
temperatures during this contingency. Similar tests were run to IX. BIOGRAPHIES
simulate the normal (noncontingency) load on the transformer. Walter A. Castillo received his B.S. in 2006 from Louisiana State University.
Fig. 5 also illustrates the algorithm responses obtained during He worked at PPG Aerospace in Huntsville, Alabama, as a process engineer
for less than a year. In 2006, he joined CenterPoint Energy and is presently a
the noncontingency situation. It is significant to note the lead engineer in the substation protection and automation group. He is a
algorithm responses between 13:35 and 15:35. The loadable registered professional engineer in the state of Texas.
MVA during contingency is observed to be less than the
loadable MVA under normal operating conditions. This is Rekha T. Jagaduri received her B.S. in 2003 from Madras University in
India and her M.S. (magna cum laude) in power systems from Tennessee
because the transformer is burdened for an extended time
Technological University. She worked as a power system engineer in the
interval under contingency and, as a result, its temperatures protection and control division for the EPC Company in Knoxville,
increase. The time constant for the top-oil temperature is Tennessee. Rekha has utility company and research experience. She was
approximately six to nine hours, and consequently, it takes involved in electric power engineering protection and control design, solution
development, and detailed engineering design. In 2008, Rekha joined
time for the temperatures to cool down, during which time the Schweitzer Engineering Laboratories, Inc. She has led large industrial
loading capability of the transformer is reduced. projects, is a member of IEEE, and has authored various technical papers. She
is a registered professional engineer in the state of Texas.
VII. CONCLUSION
Prasanna K. Muralimanohar received his B.S. in 2008 from Anna
CenterPoint Energy is presently using a time-based method University in India and his M.S. in electric power systems from Rensselaer
to monitor its distribution transformers. The shortcomings of Polytechnic Institute in 2009. His research encompassed power system
modeling and analysis. Prasanna worked as a junior protection engineer at
this method are that the load profile data used to determine the United Electric Systems in Allentown, Pennsylvania, where he was involved
transformer loading capability are based on a typical in performing protection settings analysis for industrial protection schemes. In
CenterPoint Energy example load profile and not the actual 2010, he joined Schweitzer Engineering Laboratories, Inc. as a protection
engineer.
load profile, the ambient temperatures are not the actual
values, and the inspection and analysis are performed on a
yearly basis. With the real-time monitoring solution,
CenterPoint Energy can detect transformer loading capability
abnormalities and selectively schedule maintenance. In
addition to providing CenterPoint Energy with an accurate
estimate of the loading capability of its transformers, this
solution will also help CenterPoint Energy better allocate its
maintenance crew. CenterPoint Energy intends to use this
custom-developed solution to better manage its substation
assets by considerably increasing the mean time between
maintenance, avoiding unexpected outages due to transformer
failures, and, more importantly, reducing the risk of exposing
the transformer to abnormal operating conditions.

VIII. REFERENCES
[1] M. Arshad and S. M. Islam, “Power Transformer Condition Monitoring
and Assessment for Strategic Benefits,” proceedings of the Australasian
Universities Power Engineering Conference, Christchurch, New
Zealand, September 2003.
[2] W. H. Tang and Q. H. Wu, Condition Monitoring and Assessment of
Power Transformers Using Computational Intelligence, Springer,
London, 2011, pp. 1–11.
[3] J. H. Provanzana, P. R. Gattens, W. H. Hagman, H. R. Moore,
J. W. Harley, and J. E. Triner, “Transformer Condition Monitoring –
Realizing an Integrated Adaptive Analysis System,” CIGRE 1992,
Ref. No. 12-105.
[4] W. Knorr, T. Liebfried, K. Viereck, D. Dohnal, U. Sundermann, and A.
Kosmata, “On-Line Monitoring of Power Transformers – Trends, New
Developments, and First Experiences,” CIGRE 1998, Ref. No. 12-211.
[5] I. J. Kemp, “Partial Discharge Plant-Monitoring Technology: Present Previously presented at the 2012 Texas A&M
and Future Developments,” IEE Proceedings – Science, Measurement Conference for Protective Relay Engineers.
and Technology, Vol. 142, No. 1, January 1995, pp. 4–10. © 2012 IEEE – All rights reserved.
20120217 • TP6551-01